Memento on Eviews output

Jonathan Benchimol

10th June 2013

Abstract
This small paper, presented as a memento format, reviews almost
all frequent results Eviews produces following standard econometric
studies. Although it does not include all the information that the
Eviewshelp tools include, this memento is useful in order to obtain
short responses to simple econometric output. This version has been
updated in June 2015, and started in February 2008.

Keywords: Eviews, Econometrics, Statistics.

JEL Classication: E31, E51, E58.

Bank of Israel and EABCN, POB 780, 91007 Jerusalem, Israel. Phone: +972-2-
6552641. Fax: +972-2-6669407. Email: jonathan.benchimol@boi.org.il. The content of
this paper does not necessarily reect the views of the Bank of Israel.

1
1 Introduction
This memento is intended to become a useful work tool for Eviewsusers. It
has been used by researchers in various elds such as in Economics (Benchi-
mol, 2011, 2014, 2015, 2016; Benchimol and Fourans, 2012, 2016), Oper-
ating and Maintenance (Chia, 2010), Finance (Bekale, 2015), and Energy
(Bakhtiari et al., 2015). Although it is still incomplete, it is very helpful.

2 Unit Root Tests

2.1 ADF test statistic
Augmented Dickey-Fuller

2.2 PP test statistic

Phillips-Perron

2.3 KPSS test statistic

2.4 ADF test equation

3 Generalized Method of Moments

The starting point of Generalized Method of Moments (GMM) estimation
is a theoretical relation that the parameters should satisfy. The idea is to
choose the parameter estimates so that the theoretical relation is satised
as closely as possible. The theoretical relation is replaced by its sample
counterpart and the estimates are chosen to minimize the weighted distance
between the theoretical and actual values. GMM is a robust estimator in
that, unlike maximum likelihood estimation, it does not require information
of the exact distribution of the disturbances. In fact, many common estima-
tors in econometrics can be considered as special cases of GMM.
The theoretical relation that the parameters should satisfy are usually
orthogonality conditions between some (possibly nonlinear) function of the
parameters f ( ) and a set of instrumental variables zt :

E f ( )0 Z = 0 (1)

where are the parameters to be estimated. The GMM estimator selects

parameter estimates so that the sample correlations between the instruments

2
and the function f are as close to zero as possible, as dened by the criterion
function:
J ( ) = (m ( ))0 Am ( ) (2)
where m ( ) = f ( )0 Z and A is a weighting matrix.

3.1 J-statistic
The J-statistic is the minimized value of the objective function, where we re-
port (2) divided by the number of observations. This J-statistic may be used
to carry out hypothesis tests from GMM estimation. A simple application
of the J-statistic is to test the validity of overidentifying restrictions. Un-
der the null hypothesis that the overidentifying restrictions are satised, the
J-statistic times the number of regression observations is asymptotically 2
with degrees of freedom equal to the number of overidentifying restrictions.
If the equation excluding suspect instruments is exactly identi-
ed, the J-statistic will be zero.

3.2 Coe cient of Determination

The Coe cient of Determination (R2 ) is a statistic that will give some in-
formation about the goodness of t of a model. In regression, the coe cient
of determination is a statistical measure of how well the regression line ap-
proximates the real data points. A R2 value of 1:0 indicates that the
regression line perfectly ts the data. Its often a suspicious result. As
presented in Table 1, an acceptable value for R2 is superior to 0:5.

3.3 Adjusted Coe cient of Determination

The Adjusted Coe cient of Determination (Adjusted R2 ) is a modication
of R2 that adjusts for the number of explanatory terms in a model. Unlike
R2 , the Adjusted R2 increases only if the new term improves the model more
than would be expected by chance. The Adjusted R2 can be negative (in very
poorly specied regression equations.), and will always be less than or equal
to R2 . Adjusted R2 does not have the same interpretation as R2 . As such,
care must be taken in interpreting and reporting this statistic. Adjusted R2 is
particularly useful in the feature selection stage of model building. Adjusted
R2 is not always better than R2 : adjusted R2 will be more useful only if the
R2 is calculated based on a sample, not the entire population. For example,
if our unit of analysis is a state, and we have data for all counties, then
Adjusted R2 will not yield any more useful information than R2 .

3
3.4 Mean Dependent Variable
The value of the Mean Dependent Variable is the mean of the observations
of the dependent variable.

3.5 S.D. Dependent Variable

The value of the S.D. Dependent Variable is the estimated standard deviation
of the dependent variable.

3.6 S.E. of Regression

The S.E. of Regression is a summary measure of the size of the equations
errors. The unbiased estimate of it is calculated as the square root of the
sum of squared residuals divided by the number of usable observations minus
the number of regressors (including the constant). This measure should
be closer to zero.

3.7 Sum of Squared Residual

The residual sum of squares (RSS) is the sum of squares of residuals. It is
the discrepancy between the data and our estimation model. As smaller
this discrepancy is, better our estimation will be.

3.8 Prob(F-statistic)
To test the success of the regression model, a test can be performed on R2 .
Usually, we accept that the regression model is useful when the
Prob(F-statistic) is smaller than the desired signicance level, for
example 0.05 (for 5% signicance level).

3.9 Durbin-Watson statistic

The Durbin-Watson statistic is a test statistic used to detect the presence of
autocorrelation in the residuals from a regression analysis. Its value always
lies between 0 and 4.
A value of 2 indicates there appears to be no autocorrelation. If the
Durbin-Watson statistic is substantially less than 2, there is evidence of pos-
itive serial correlation and values much above 2 are indicative of negative
serial correlation. As a rough rule of thumb, if Durbin-Watson statistic is
less than 1.0, there may be cause for alarm. Small values of Durbin-Watson

4
statistic indicate successive error terms are, on average, close in value to one
another, or positively correlated. Large values of Durbin-Watson statistic in-
dicate successive error terms are, on average, much dierent in value to one
another, or negatively correlated. How much below or above 2 is required for
signicance depends on the number of usable observations and the number
of independent variables (excluding the constant).
The Durbin-Watson test is a test for rst-order serial correlation in the
residuals of a time series regression. A value of 2.0 for the Durbin-
Watson statistic indicates that there is no serial correlation but this
result is biased toward the nding that there is no serial correlation
if lagged values of the regressors are in the regression.

3.10 Determinant residual covariance

The Determinant residual covariance is the determinant of the residual co-
variance matrix. If the determinant of the residual covariance matrix
is zero, the estimates are e cient. But, if a comparison of two determi-
nants of eachs residual covariance matrix shows a value, for example, >100
for the original VAR and a value near to zero for the log-VAR, then a lin-
early dependent covariance matrix seems unlikely, the zero-value must be
due to very small covariances (but these are caused by the transformation
into log-units, and must not be due to a real improvement of the model).

4 Maximum likelihood
Maximum Likelihood Estimation (MLE) is a popular statistical method used
to calculate the best way of tting a mathematical model to some data.
Modeling real world data by estimating maximum likelihood oers a way of
tuning the free parameters of the model to provide an optimum t.
The likelihood and log-likelihood functions are the basis for deriving esti-
mators for parameters, given data. While the shapes of these two functions
are dierent, they have their maximum point at the same value. In fact, the
value of p that corresponds to this maximum point is dened as the Maxi-
mum Likelihood Estimate (MLE). This is the value that is mostly likely"
relative to the other values. This is a simple, compelling concept and it has
a host of good statistical properties.

5
4.1 Log likelihood
The shape of the log-likelihood function is important in a conceptual way. If
the log-likelihood function is relatively at, one can make the interpretation
that several (perhaps many) values of p are nearly equally likely. They are
relatively alike. This is quantied as the sampling variance or standard
error. If the log-likelihood function is fairly at, this implies considerable
uncertainty and this is reected in large sampling variances and standard
errors, and wide condence intervals.
On the other hand, if the log-likelihood function is fairly peaked near its
maximum point, this indicates some values of p are relatively very likely com-
pared to others. There is some considerable degree of certainty implied and
this is reected in small sampling variances and standard errors, and narrow
condence intervals. So, the log-likelihood function at its maximum
point is important as well as the shape of the function near this
maximum point.

4.2 Avg. log likelihood

Average log likelihood is the log likelihood (i.e. the maximized value of
the log likelihood function) divided by the number of observations. The
maximization of the log-likelihood is the same as the maximization
of the average log likelihood. This statistic is useful in order to compare
models.

4.3 Akaike Information Criterion

Akaikes Information Criterion (AIC) is a measure of the goodness of t of
an estimated statistical model. It is grounded in the concept of entropy.
The AIC is an operational way of trading o the complexity of an estimated
model against how well the model ts the data.
The preferred model is the one with the lowest AIC value. The
AIC methodology attempts to nd the model that best explains the data with
a minimum of free parameters. By contrast, more traditional approaches to
modeling start from a null hypothesis. The AIC penalizes free parameters
less strongly than does the Schwarz criterion.

4.4 Schwarz Information Criterion

The Bayesian information criterion (BIC) is a statistical criterion for model
selection. The BIC is sometimes also named the Schwarz criterion, or Schwarz

6
information criterion (SIC). It is so named because Gideon E. Schwarz (1978)
gave a Bayesian argument for adopting it.
Given any two estimated models, the model with the lower value
of BIC is the one to be preferred. The BIC is an increasing function
of residual sum of squares and an increasing function of the number of free
parameters to be estimated (for example, if the estimated model is a linear
regression, it is the number of regressors, including the constant). That is,
unexplained variation in the dependent variable and the number of explana-
tory variables increase the value of BIC. Hence, lower BIC implies either fewer
explanatory variables, better t, or both. The BIC penalizes free parameters
more strongly than does the Akaike information criterion.

4.5 Hannan-Quinn Information Criterion

Ideally AIC and SBIC should be as small as possible (note that all can
be negative). Similarly, the Hannan-Quinn Information Criterion
(HQIC) should be also as small as possible. Therefore the model to be
chosen should be the one with the lowest value of information criteria test.

4.6 Determinant residual covariance

Maximizing the likelihood value is equivalent to minimizing the determinant
of the residual covariance matrix. Thus, the determinant of the residual
covariance matrix and not the residuals itself are minimized. As smaller
this determinant is, better our estimation will be.

5 Summary table

7
 Type Optimal Acceptable
R and Adjusted R2
2
!1 > 0:5
J-statistic !0 < 0:1
Mean dependant variable ! +1 > 100
S.E. of Regression !0 Choose the lower value (comparison)
Residual sum of squares !0 Choose the lower value (comparison)
Prob(F-statistic) !0 < 0:05
Durbin-Watson statistic !2 1:8 < DW < 2:2 (Under conditions)
Determinant residual covariance !0 Choose the lower value (comparison)
3
Log-Likelihood ! +1 > 10
Average Log-Likelihood ! +1 > 10
AIC ! 1 Choose the lower value (comparison)
SIC ! 1 Choose the lower value (comparison)
HQIC ! 1 Choose the lower value (comparison)

Table 1: Summary table

8
References
Bakhtiari, A. A., Hematian, A., Shari, A., 2015. Energy analyses and green-
house gas emissions assessment for saron production cycle. Environmental
Science and Pollution Research, 118.

Bekale, A. N., 2015. Institutionalisation of derivatives trading and economic

growth : evidence from South Africa. Master dissertation, University of
South Africa.

Benchimol, J., 2011. New Keynesian DSGE models, money and risk aversion.
PhD dissertation, UniversitParis 1 Panthon-Sorbonne.

Benchimol, J., 2014. Risk aversion in the Eurozone. Research in Economics

68 (1), 3956.

Benchimol, J., 2015. Money in the production function: a New Keynesian

DSGE perspective. Southern Economic Journal 82 (1), 152184.

Benchimol, J., 2016. Money and monetary policy in Israel during the last
decade. Forthcoming in Journal of Policy Modeling.

Benchimol, J., Fourans, A., 2012. Money and risk in a DSGE framework: a
Bayesian application to the Eurozone. Journal of Macroeconomics 34 (1),
95111.

Benchimol, J., Fourans, A., 2016. Money and monetary policy in the Euro-
zone: an empirical analysis during crises. Forthcoming in Macroeconomic
Dynamics.

Chia, T., 2010. A model to estimate the operating and maintenance costs of
the mine resistant ambush protected vehicles. Master dissertation, Naval
Postgraduate School.